Lavapipe is CPU rendering, it doesn't really prove much. But also, Vulkan on BSDs is totally possible and isn't something esoteric, FreeBSD has it.
> Build goal only: This targets compilation and linkage of the Vulkan stack. Runtime GPU acceleration is not available under VirtualBox; the software driver (Lavapipe) is the target.
I don't understand why this would ever be a problem, even without LLM assistance it's something that sounds like a weekend project?
this is what i would recommend. hoist it over from freebsd. it works well vulkan is fully usable, mesa works nicely. Even seen people playing with CUDA tho i didnt get that workin myself yey.
- Vulkan as a runtime service: 'the ability' for programs to do Vulkan things
- Vulkan as an API / specification: how programs interact with the service to do those things, or how the service interacts with programs
- Vulkan as a technology: how the API calls are implemented
Generally Vulkan refers to the specification much like how Linux refers to the kernel, but there is still Linux (kernel) vs Linux (syscall ABI) vs Linux (runtime service)
Actually, it's orthodox; and it's fetch that isn't. FreeBSD is actually the odd one out, for having an extra tool for doing the same thing. The ftp tool in all of the BSDs, including FreeBSD, speaks HTTP, and has done since Luke Mewburn did lukemftp (later to be named tnftp) and Theo de Raadt did likewise, both based on the original 4.2BSD ftp, back in the middle 1990s.
Nice! Thank you, that explains it. FreeBSD was my first *BSD and the only one I had long-term contact with, so I unconsciously use it as my "BSD standard". My bad on that :)
I dunno, that feels very BSD to me. Presumably, they had a ftp utility first, and then when somebody wanted to download files over http they looked around and decided that the obvious thing to do was to add it to the existing file transfer/download program. Same as continuing to add functions to ifconfig rather than inventing a new ip tool.
Sure, if those functions added were related to configuring network interfaces, but it would be odd if someone added functions for configuring storage controllers.
The sticking point is always what "one thing" is. I can certainly see an argument for "download a file over http" (or, for another program, ftp), but I can also see an argument for "download files over the internet". In which case, honestly the only reason I'd even quirk an eyebrow at a BSD's ftp(1) adding bittorrent support is that bittorrent usually means running a daemon that that does feel out of character for it. But even then I wouldn't be that surprised to see it added.
Oldest supported machine for NetBSD is VAX 780 from 1978(!!!). One of the first system supporting mmu, 32 bit cpu, virtual memory etc etc
This machine is so slow that it takes a lot of time to generate ssh keys etc. We talking here hours hehe
NetBSD is known to support like 60 architectures - many of them low end embedded systems: so ftp AS A CHOICE (you have other options!) is very smart and easy
I think GP is confused why the ftp command also handles http(s) :)
I hate to imagine what a 780 running NetBSD would be like, too.
I tried netbooting NetBSD on my MicroVAX 3400, which is about 2.5x the performance of the 780. It did, literally, take 6+ hours to slog through making RSA keys.
> I think GP is confused why the ftp command also handles http(s) :)
Exactly - I even suspected for a second that `ftp` on NetBSD is something else entirely, not an actual FTP client with HTTP/HTTPS URLs bolted on. It's not - it still accepts a host as an argument and opens a CLI if there's an FTP server to talk to.
Yep, it's a loooot faster. But then, when you decide to boot NetBSD on these boxes, you're not really doing it for practical reasons anyway, so part of the experience is waiting a few hours for /etc/rc.d/sshd to do its thing.
Probably more relevant on more "borderline" hosts. My SPARCstation 2 can just barely run NetBSD 10 at what I'd call "tolerable speeds" for some concept of "real work", and it's something like 50x the speed of the 780!
The vulkan stack is rather lean (there are still c++ though, valve removed a lot of c++ for less c++, it would have been correct with plain and simple C).
The big chunk is DRM kernel code.
AMD seems to be working on _userland_ hardware command ring buffers, which should makes userland vulkan even simpler. Dunno how they will work around the VMID stuff though.
I think both points are valid; that is the sensible way to start, but it's worth noting that the work is early so people don't misunderstand the current state of things.
Lavapipe is CPU rendering, it doesn't really prove much. But also, Vulkan on BSDs is totally possible and isn't something esoteric, FreeBSD has it.
> Build goal only: This targets compilation and linkage of the Vulkan stack. Runtime GPU acceleration is not available under VirtualBox; the software driver (Lavapipe) is the target.
I don't understand why this would ever be a problem, even without LLM assistance it's something that sounds like a weekend project?
this is what i would recommend. hoist it over from freebsd. it works well vulkan is fully usable, mesa works nicely. Even seen people playing with CUDA tho i didnt get that workin myself yey.
FreeBSD relies on being able to run the Linux graphics drivers though via a compat layer.
NetBSD uses the Linux graphics drivers too.
> Vulkan is now available
looks inside:
> What this is NOT (yet): Running Vulkan programs
Technically, it isn't wrong, if you consider Vulkan the technology rather than the service
isn't Vulkan the api?
'Vulkan' can refer to at least:
- Vulkan as a runtime service: 'the ability' for programs to do Vulkan things
- Vulkan as an API / specification: how programs interact with the service to do those things, or how the service interacts with programs
- Vulkan as a technology: how the API calls are implemented
Generally Vulkan refers to the specification much like how Linux refers to the kernel, but there is still Linux (kernel) vs Linux (syscall ABI) vs Linux (runtime service)
Installation instructions:
That's... a bit unorthodox. FreeBSD has a `fetch`[1] utility for this, I wasn't aware NetBSD puts that in `ftp`[2].Interesting choice. I wonder what led to it.
[1] https://man.freebsd.org/cgi/man.cgi?fetch
[2] https://man.netbsd.org/ftp.1
Actually, it's orthodox; and it's fetch that isn't. FreeBSD is actually the odd one out, for having an extra tool for doing the same thing. The ftp tool in all of the BSDs, including FreeBSD, speaks HTTP, and has done since Luke Mewburn did lukemftp (later to be named tnftp) and Theo de Raadt did likewise, both based on the original 4.2BSD ftp, back in the middle 1990s.
* https://cgit.freebsd.org/src/tree/contrib/tnftp/ChangeLog#n1...
* https://cvsweb.netbsd.org/bsdweb.cgi/src/usr.bin/ftp/main.c?...
* https://cvsweb.openbsd.org/log/src/usr.bin/ftp/main.c,v?sort...
Nice! Thank you, that explains it. FreeBSD was my first *BSD and the only one I had long-term contact with, so I unconsciously use it as my "BSD standard". My bad on that :)
I dunno, that feels very BSD to me. Presumably, they had a ftp utility first, and then when somebody wanted to download files over http they looked around and decided that the obvious thing to do was to add it to the existing file transfer/download program. Same as continuing to add functions to ifconfig rather than inventing a new ip tool.
Sure, if those functions added were related to configuring network interfaces, but it would be odd if someone added functions for configuring storage controllers.
Imagine using an ftp program to transfer files.
BitTorrent transfers files, so should a BT client be bolted on too? Typically the Unix philosophy has been do one thing well, not do all the things.
The sticking point is always what "one thing" is. I can certainly see an argument for "download a file over http" (or, for another program, ftp), but I can also see an argument for "download files over the internet". In which case, honestly the only reason I'd even quirk an eyebrow at a BSD's ftp(1) adding bittorrent support is that bittorrent usually means running a daemon that that does feel out of character for it. But even then I wouldn't be that surprised to see it added.
Oldest supported machine for NetBSD is VAX 780 from 1978(!!!). One of the first system supporting mmu, 32 bit cpu, virtual memory etc etc
This machine is so slow that it takes a lot of time to generate ssh keys etc. We talking here hours hehe
NetBSD is known to support like 60 architectures - many of them low end embedded systems: so ftp AS A CHOICE (you have other options!) is very smart and easy
I think GP is confused why the ftp command also handles http(s) :)
I hate to imagine what a 780 running NetBSD would be like, too.
I tried netbooting NetBSD on my MicroVAX 3400, which is about 2.5x the performance of the 780. It did, literally, take 6+ hours to slog through making RSA keys.
> I think GP is confused why the ftp command also handles http(s) :)
Exactly - I even suspected for a second that `ftp` on NetBSD is something else entirely, not an actual FTP client with HTTP/HTTPS URLs bolted on. It's not - it still accepts a host as an argument and opens a CLI if there's an FTP server to talk to.
maybe rather than protocol its program. then it all makes sense no. File Transfer Program. voila.
Elliptic curve key gen is much faster on slow hardware. You're waiting around on primality tests that aren't necessary with modern keys.
Yep, it's a loooot faster. But then, when you decide to boot NetBSD on these boxes, you're not really doing it for practical reasons anyway, so part of the experience is waiting a few hours for /etc/rc.d/sshd to do its thing.
Probably more relevant on more "borderline" hosts. My SPARCstation 2 can just barely run NetBSD 10 at what I'd call "tolerable speeds" for some concept of "real work", and it's something like 50x the speed of the 780!
Disable some booted services at /etc/rc.conf:
https://luke8086.dev/netbsd-on-thinkpad-380z.html
Easier with SIMH.
This is a nice project but looks like is either AI written or AI assisted and I haven’t seen mention of that in any of the docs.
Didn't modular-xorg, MESA and DRM drivers handle this?
I expected this to be official from the title but it doesn’t seem to be.
The vulkan stack is rather lean (there are still c++ though, valve removed a lot of c++ for less c++, it would have been correct with plain and simple C).
The big chunk is DRM kernel code.
AMD seems to be working on _userland_ hardware command ring buffers, which should makes userland vulkan even simpler. Dunno how they will work around the VMID stuff though.
There are already Vulkan components in pkgsrc and wip.
Which ones? I'm asking, because I wouldn't have done anything at all if this was pre-existing.
See graphics/MesaLib.
I have never had a need for NetBSD, but in case I ever do, I’m glad it’s there. Especially with Linux deprecating old platforms.
This looks like an unofficial effort but hopefully it gets refined and integrated.
NetBSD already has relevant packages officially.
Which packages? I couldn't find what I needed, and that's the whole reason for building all this.
Lavapipe? So it's just Mesa software rendering stuff
That's the start. The idea is to get Vulkan running first. More to follow.
I think both points are valid; that is the sensible way to start, but it's worth noting that the work is early so people don't misunderstand the current state of things.
This is 100% AI slop.